I got this problem:

Prove that there is no function $f:\Bbb{R}\to\Bbb{R}$ with $f(0)>0$ such that $\forall x,y\in\Bbb{R}, f(x+y)\geq f(x)+y f(f(x))$.
(Hint: the solution involves limits at infinity)

I tried to prove it (by contradiction) but I failed. Thanks on any help.


DISCLAIMER: This answer is incorrect as it stands, namely in the $k>0$ section is claims that: $$ f(0)=f(x-x)\geq f(x)-x f(f(x)) $$ implies $f(x)<f(0)$, which is false.

This disclaimer will stay here until the problem is fixed.

From the condition $f(0)>0$ it follows that with $x=0$ we have $$ f(0+y)-y f(f(0))\geq f(0)>0 $$ so it follows that $f(y)>ky$ where $k=f(f(0))$ is some non-zero constant. It has to be non-zero since otherwise $0=k=f(f(x))>k f(x)=0$. Now we have to divide into cases.

If $k>0$

If $k>0$ it follows that $x f(f(x))>kx f(x)>k^2 x^2\geq 0$. So in that case we have $$ \underbrace{f(x-x)}_{f(0)}\geq f(x)-\underbrace{x f(f(x))}_{\text{strictly greater than }0}\\ \iff\\ f(x)<f(0) $$ But the above statement is absurd by all means. For one thing $f(0)<f(0)$ and moreover $f(0)>f(x)>kx\rightarrow\infty$ which is also absurd.

If $k<0$

If we apply the beginning of what Denis showed, namely that $f(x)\rightarrow\infty$ for $x\rightarrow\infty$, then we get in addition to that from my analysis that for $k<0$ we have $f(-x)\geq -kx\rightarrow\infty$ for $x\rightarrow\infty$. So we see that $$ f(0)=f(-x+x)\geq f(-x)+x f(f(-x))\rightarrow\infty $$ for $x\rightarrow\infty$ too. A contradiction since we have just shown that the constant $f(0)$ is greater than or equal to an expression that tends to infinity. So $k<0$ is impossible too.

  • $\begingroup$ If you put x = 0, then the LHS in the first inequality you gave should be f(0) - yf(f(0)) $\endgroup$ – TenaliRaman Sep 24 '14 at 8:43
  • 1
    $\begingroup$ @TenaliRaman: No! The LHS is still $f(0+y)-y f(f(0))$ even if $y=0$, and the statement $f(y)>ky$ holds for ANY $y$ including $y=0$. $\endgroup$ – String Sep 24 '14 at 8:49
  • $\begingroup$ @TenaliRaman: But I spotted another serious problem in my analysis which I just pointed out in my answer above ... $\endgroup$ – String Sep 24 '14 at 8:53
  • 1
    $\begingroup$ @String: Maybe we can suppose that $\forall x\in\Bbb{R}, f(f(x))\leq 0$ and from that reach to a contradiction and get that $\exists z\in\Bbb{R}$ such that $f(f(z))>0$ which will be something like the $k$ that you mentioned. $\endgroup$ – MathNerd Sep 24 '14 at 9:04
  • 1
    $\begingroup$ @Saita: I finished my argument! $\endgroup$ – String Sep 24 '14 at 10:32

Let $a=f(0)>0$, $b=f(a)$ and $c=f(b)$.

We know that $f(0+b)\geq a+b^2$, so we get $c>0$.

We also have for all $x$, $f(x+a)\geq f(a)+cx$, so in particular $\lim_{x\to\infty} f(x)=\infty$ and more precisely $f(x)=\Omega(x)$ when $x\to\infty$.

This means that for $x$ big enough, $f(x)>0$, and therefore, $f(x+1)\geq f(x)+f(f(x))>f(f(x))$. This gets us $f(f(x-1))<f(x)$.

Moreover, for $x$ big enough, we have $f(x-1)>x$ (since $f(x+1)=\Omega(x^2)$).

Let us choose $x_0$ big enough, and define $x_{n+1}=f(x_n-1)$.

The sequence $(x_n)$ is strictly increasing, and for all $n$, we have $f(x_{n+1})<f(x_n)$. We reached a a contradiction with $\lim_{x\to\infty} f(x)=\infty$.

  • $\begingroup$ So to sum up you have shown $f(x+a)\geq cx+b$ where $c>0$. Very nice! How do you conclude from that? $\endgroup$ – String Sep 24 '14 at 9:29
  • $\begingroup$ I found the end :) $\endgroup$ – Denis Sep 24 '14 at 10:11
  • $\begingroup$ It took me a while to absorb, but it makes sense now. Perhaps you should state the inequalities as $f(x)>f(f(x-1))$ and $f(x-1)>x$ respectively to make the reading more direct, but that is just an opinion. The first part of your answer showing that $f(x)\rightarrow\infty$ provided the last piece of the puzzle for my version of it BTW :) $\endgroup$ – String Sep 24 '14 at 10:40
  • 1
    $\begingroup$ I smoothed it accordingly. $\endgroup$ – Denis Sep 24 '14 at 10:47

By what Denis said we get that $lim_{x\to\infty}f(x)=\infty$ And so $lim_{x\to\infty}f(f(x))=\infty$.

Now we will reach a contradiction by using the definition of limits at infinity:

Because $lim_{x\to\infty}f(x)=\infty$ and $lim_{x\to\infty}f(f(x))=\infty$. We get that:
$\forall 0<M_1,\exists 0<N_1$ such that $\forall N_1<x, M_1<f(x)$ and that
$\forall 0<M_2,\exists 0<N_2$ such that $\forall N_2<x, M_2<f(f(x))$

Now take $M_1=M_2=1$ and we will get that $\exists 0<N_1$ such that $\forall N_1<x, 1<f(x)$ and that $\exists 0<N_2$ such that $\forall N_2<x, 1<f(f(x))$

Take $x=max (N_1+1,N_2+1)$, then $N_1,N _2<x$ and so we get that $1<f(x)$ and that $1<f(f(x))$

Now take $y=max(\frac{x+1}{f( f( x))-1},N_2-x)$ which implies that $N_2< x+y+1\leq y f(f(x ))$ and so $1<f(f(x+y+1))$, Now $f (x+y)\geq f(x )+yf(f(x))\geq x+y+1$ And hence

$f(f(x+y))\geq f(x+y+1)+(f(x+y)-(x+y+1))f(f(x+y+1)) \geq f(x +y+1)\geq f(x+y)+f(f(x+y))\geq f(x)+yf(f(x))+f(f(x+y))>f(f(x+y))$

And so we get that $0>0$ which is a contradiction.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.